On Fri Sep 13 22:39:18 2024 +0000, Elizabeth Figura wrote:
Neither is the sync_ioctl() IOSB. The problem is that for user I/O the 32-bit iosb will always exist, so our I/O infrastructure assumes it will and tries to write to it. Making it not do that is a few orders of magnitude harder than just supplying the dummy iosb.
I see, thanks, I will update that now (along with avoiding set_fd_cancel, probably then renaming no_cancel to is_system in async structure.